Gridding global male and female populations: New data from the

Gridded Population of the World (GPW)

Erin Doxsey-WhitfieldSusana B. AdamoKytt MacManus

November 10, 2015EFGS – Vienna, Austria

http://sedac.ciesin.columbia.edu

2

Overview

•What is GPW?•Methods for creating GPWv4 grids of population by sex• Challenges in acquiring and integrating data• Early look at grids of males and females

Gridded Population of the World (GPW)• Raster data product developed to provide a spatially-disaggregated population

surface that is compatible with data sets from social, economic, and Earth science fields.

• Census population data are transformed from their native spatial units to a global grid of quadrilateral latitude-longitude cells (Balk et al. 2010)

• Free and openly available

GPW version 3, 2000 population density Transforming census units to a grid 3

Publication Year

Years of Estimation

Grid Resolution

Number of Input Units

Census variables Population Density Grid

GPWv1 1995 1994 5 arc-minute (~10 km) 19,000 Total Population

GPWv2 2000 1990, 1995 2.5 arc-minute (~5 km) 127,000 Total Population

GPWv3 2005 1990, 1995, 2000

2.5 arc-minute (~5 km) ~ 400,000 Total Population

GPWv4 20152000, 2005, 2010, 2015,

202030 arc-second

(~1 km) ~ 12,500,000Total Population,

Sex, Age, Urban/Rural

status

Development of GPW

2010

2000

1995

1994

4

5

GPW Applications

• Broad use in research, policy making, communications, and human and environmental problem-solving

• Often combined with satellite remote sensing or other biophysical data

• Demand from research community to include demographic information in global population grids

700+ GPWv3 citations to date

6

GPW ApplicationsSustainable Development Goals

“In order to monitor the implementation of the SDGs, it will be important to improve the availability of and access to data and statistics disaggregated by income, gender, age, race, ethnicity, migratory status, disability, geographic location and other characteristics relevant in national contexts.”

-SDG Open Working Group report

GPWv4: Global population gridded by sex, age, and urban/rural designation

“In order to monitor the implementation of the SDGs, it will be important to improve the availability of and access to data and statistics disaggregated by income, gender, age, race, ethnicity, migratory status, disability, geographic location and other characteristics relevant in national contexts.”

-SDG Open Working Group report

7

GPWv4 Methods: Population grids by sex

Find tabular population counts

Hierarchy of Population Data1. Census or Population Register data2. Official total population estimates from NSO3. National-level estimates by sex for 2010 from UN URPAS

8

GPWv4 Methods: Population grids by sex

Find tabular population counts

Match to geographic boundaries (census or

administrative)

9

GPWv4 Methods: Population grids by sex

Find tabular population counts

Match to geographic boundaries (census or

administrative)

Adjust boundaries to global framework

10

GPWv4 Methods: Population grids by sex

Find tabular population counts

Match to geographic boundaries (census or

administrative)

Adjust boundaries to global framework

Estimate the population for target years (2000,

2005, 2010, 2015, 2020)

11

GPWv4 Methods: Population grids by sex

Adjust boundaries to global framework

Estimate the population for target years (2000,

2005, 2010, 2015, 2020)

Estimate male and female populations in 2010

Find tabular population counts

Match to geographic boundaries (census or

administrative)

• Calculate proportion of M and F M/(M+F) and F/(M+F)

• Apply to the 2010 estimates of total population

12

GPWv4 Methods: Population grids by sex

Adjust boundaries to global framework

Estimate the population for target years (2000,

2005, 2010, 2015, 2020)

Estimate male and female populations in 2010

Find tabular population counts

Match to geographic boundaries (census or

administrative)

Proportionally-allocate population to 1 km grids using

an areal-weighting method

13

Challenges: Availability of Sex Data

Population by sex not as readily available as total population

Use national-level estimates for 2010 from UN URPAS (19 countries)

• No recent census: e.g. Lebanon, Iraq, Somalia, Uzbekistan

• Challenges in finding tabular data: e.g. Morocco

Solution 1 Solution 2Use coarser resolution data for sex

(16 countries)• Canada: too much data suppression at L5;

used L3 for more complete coverage• Colombia: 2005 census was adjusted

following it’s release; however adjusted sex data was not available to the same level as total pop

Calculate proportions of males and females at coarse level, apply to higher resolution 2010 total population estimate

14

Challenges: Privacy constraints

Privacy Constraints

Random rounding

Suppression of characteristic data for units with high non-response rates

Suppression of characteristic data

for units with counts below a threshold

Implications:• Choose sex data from a coarser

resolution• Male grid + Female grid ≠ Total grid• Grids may have No Data for a

number of units

15Preliminary data from GPWv4, 2010 estimate

16Preliminary data from GPWv4, 2010 estimate

17Preliminary data from GPWv4, 2010 estimate

18Preliminary data from GPWv4, 2010 estimate

19

PreliminarySex Ratio Grids

India Mozambique

Preliminary data from GPWv4, 2010 estimate

20

Evaluation of Census Data QualityGPWv4 will include national-level metadata and other qualitative indicators to inform users of:

1. The currency of the population data 2. Whether the data is from estimates, population

registers, preliminary censuses, or final censuses3. Coverage errors evaluated in

post-enumeration surveys4. Evaluation of age-sex distribution

Census Quality

Interpreta-bility

Coherence

Accessi-bility

Accuracy

Timeliness

Relevance

21

• Provide classic numeric indices of age-sex quality• Whipple’s index• Myer’s Blended Method• UN Age-sex accuracy index

• Provide graphs• Provide text describing

results

Mozambique 2007 Population Census: Age-Sex Distribution

Age heaping at ages ending in 0 and 5

Evaluation of Census Data Quality:Age-Sex Distribution

Under-enumeration

Overall, GPW is a census-based global population grid that is:

Email gpw@ciesin.columbia.edu to request:• Grids of total population (for beta testing)• Grids of population by sex (preliminary data)

Erin Doxsey-Whitfieldedwhitfi@ciesin.columbia.edu

• Highly used• Widely applicable

• Free and open source• Well-documented and

transparent

• Expanding to include global grids of sex, age, and urban/rural designation

22